Deep Learning Approach Versus Traditional Machine Learning for Adhd Classification

Abstract

Magnetic resonance is the imaging method that stands out in the evaluation of textures and diseases related to brain. The information about metabolic, biochemical and hemodynamic structure of the brain is obtained by magnetic resonance imaging. Attention Deficit Hyperactivity Disorder (ADHD) is a psychiatric disease and, if not treated, its effects may spread over all lifetime and cause significant academic, social, and psychiatric problems. High-accuracy and objective tools need to be developed for classification of ADHD. In this study, we present machine learning (ML) and deep learning (DL) based approaches for the classification of MR Images collected from ADHD patients. We generate a new 2D texture from 3-D structural magnetic resonance image by combining slices where gray and white matter clearly displayed. In the first approach, we extract Haralick texture based features, and HOG features and classify ADHD using ML methods such as Decision Tree, K nearest neighbor, Naive Bayes, Logistic Regression, and Support Vector Machine. In the DL approach, we trained four Convolutional Neural Network (CNN) structures (AlexNet, VGGNet, ResNet and GoogleNet) for ADHD classification using the 2-D texture images. Classification performance obtained with ResNet architecture in characterizing new texture is 100 % accuracy, 100 % sensitivity, 100 % specificity.